On Sat, Apr 19, 2014 at 5:47 PM, Eli Zaretskii <e...@gnu.org> wrote: > That's better, IMO. I think we already use ^? for different purposes > (in the Tag Table), though. Not sure if that matters, but see below. >
^? characters would only be inter >> I.e., no quoting of file names at all, and quoting node names in ^? >> (0xFF) when they contain forbidden characters. I don't think this >> looks too bad. > > ^? is 0x7F, not 0xFF. > >> E.g., >> "File: quoting.info, Node: Top, Prev: Normal 1, Next: ^?std::cout^?, >> Up: Normal 2" >> >> for a node line. > > Did you try this with Tag Tables? A tag table looks like this: ^_ Tag Table: Node: Top^?89 Node: Ch1^?292 Ref: Overview-Footnote-1^?30045 ^_ End Tag Table The quoting could be like this: ^_ Tag Table: Node: Top^?0 Node: ^?std::cout^?98 Node: Normal 1^?178 Node: Normal 2^?304 ^_ End Tag Table This particular case worked well - I could still get to the other nodes with "g" (see attachment), although it is conceivable that there could be problems, depending on what characters appear in the node/anchor name. If problems do occur with this, one solution would be to have a second tag table section in the file. > It's okay not to solve the cases where an Info file use ^_ (I don't > think you will find any of these, for this very reason). But I don't > think we can just disregard the control characters that do no harm > today, because an Info reader that will try interpreting them as a > quote might do some very wrong things with them, or even crash. IMO, > if we want the solution to stand the test of time, it should provide > some way of having "harmless" control characters in the file, without > interpreting them as quotes. The control characters would only be interpreted as quotes in particular contexts. Any outside those contexts would be left as they are. For example, if a manual contained "Type the ^? character" and the ^? was a literal ^?, it would be passed straight through, because it is not part of a node specification.
quoting.info
Description: Binary data